NICOP - Tripwire Generation for a Surgical ITS: What is the effectiveness of a data driven approach?

Abstract

Developing feedback technology for intelligent tutoring systems (ITSs) is an important and challenging problem. Current approaches require human experts to spend considerable time and effort performing task analysis and specifying tripwires for encoding within an ITS. We propose to investigate the benefits of a data driven approach for the formulation and management of ITS tripwires. In this scenario, computer algorithms will analyse simulation data from trainees and suggest potential tripwires. Such an approach can generate more complex explanations based on multivariate rules, having greater potential to explain a wider range of novice behaviour than tripwires specified by the expert alone.This is a step towards achieving higher quality feedback for novices within the ITS, based on a two-level architecture. At the lower level, tripwires will be inferred algorithmically, primed by expert knowledge and based on historical data, requiring reduced effort from human experts. At the higher level, human experts will be free to focus on providing strategic, cognitively rich feedback and explanations to novices. A key question is to investigate the most effective separation between these two levels - i.e. where is the complexity sweet spot for explanations and how feasible would it be for tripwires to provide human-like feedback? The target is to achieve greater training performance by improving feedback and quality of explanations and lowering the burden on experts in charge of intelligent tutoring systems.

Document Details

Document Type
DoD Grant Award
Publication Date
Jan 04, 2017
Source ID
N629091712033

Entities

People

  • James Bailey

Organizations

  • Office of Naval Research
  • United States Navy
  • University of Melbourne

Tags

Fields of Study

  • Computer science

Readers

  • Artificial Intelligence
  • Sensor Fusion and Tracking Systems.
  • Team-Based Human-Centered Cognitive Task Decision Making and Information Performance.

Technology Areas

  • AI & ML